The market for consumer-grade surveillance cameras in China is at a pivotal juncture.
Data from industry analysts indicates that total channel sales for 2025 reached 55.07 million units, representing a modest growth of just 3.0% year-on-year. This signals a transition from a phase of rapid market expansion to one focused on replacement within the existing user base. Concurrently, market share is increasingly consolidating around a few major brands, highlighting a pronounced oligopolistic trend.
Is There Room for New Brands?
Given this landscape of saturated growth and dominant players, is there still opportunity for new entrants? Yang Zuoxing, Chairman of Hangzhou Yanjiwei Electronics Co., Ltd. and founder of the Shenmou brand, believes there is, and his company's products offer a surprising answer. Leveraging its self-developed, ultra-low-power chips, Shenmou has carved out a niche in the battery-powered and 4G camera segment. Within a year of launch, it topped the domestic sales charts for battery cameras on major e-commerce platforms. Yang further predicts that the shift towards wireless technology will significantly catalyze the camera industry's growth, with global annual shipments potentially reaching 10 billion units by 2035.
Beyond the security market, he harbors even more ambitious goals: to deploy these proprietary low-power chips in smartphones, PCs, and servers. He contends that while Chinese companies may struggle to surpass NVIDIA in training compute power, they hold a significant opportunity to lead in the realm of inference compute.
The Untapped Potential of Security Cameras
Before entering the camera market, Yang had already built extensive experience in the semiconductor industry. His pioneering fully-custom chip design methodology had previously proven its commercial value in blockchain computing chips.
In 2020, he founded Hangzhou Yanjiwei Electronics to apply his chip design philosophy to AI semiconductors and simultaneously incubated the AI smart camera brand, Shenmou.
His foray into the security sector was driven by the identification of a fresh market opportunity. Yang pointed out that despite decades of development, the camera industry has persistently grappled with a major pain point: the power cord. He argues that since devices like phones, cars, and laptops operate without cords, a camera—essentially a visual sensor—shouldn't be tethered either. This cord, he notes, severely limits application scenarios.
For instance, installing a camera outdoors involving wiring, erecting poles, and burying cables can cost tens of thousands and take at least two months, excluding permitting. Even indoor installation requires hiring a technician for drilling and wiring, taking hours and costing hundreds of dollars—a significant expense compared to the one to two hundred dollar price tag of a consumer-grade camera itself.
Yang shared comparative data: over 5 billion mobile phone camera sensors are produced annually, whereas the figure for security cameras is only 300 to 400 million. Theoretically, IoT cameras serving all things should far outnumber phone cameras serving people, but reality has not matched this. "Installation issues constrain camera development, with the biggest problem being the cord. Solving the low-power challenge is difficult from a product perspective; it must be addressed at the chip level," he stated.
Leveraging Chip Design for Market Disruption
To tackle this industry challenge, Yang's fully-custom chip design methodology came into play again. This approach employs custom cell design, handwritten netlists, and manual layout, collectively reducing the chip's power-consumption-cost product by an order of magnitude. He applied this method to camera SoCs, image sensors, and power management chips to holistically address the device's power consumption.
It is understood that Shenmou cameras, equipped with the self-developed "Yanjixin" chip, can achieve ultra-low power consumption up to one-tenth of the industry standard. Coupled with a strategy offering free 4G data usage, this has enabled a product innovation characterized by "zero electricity costs, zero network fees, and no wiring."
Sales data shows that since commencing domestic e-commerce sales in 2023, Shenmou reached the number one position in domestic battery cameras within a year. In the first half of 2025, Shenmou also ranked first in the emerging e-commerce segments for low-power and 4G cameras. Cumulative product sales have now surpassed 10 million units.
"Shenmou cameras have more female than male users," Yang noted, explaining that while men traditionally handled installation, women are often the primary users. Shenmou's cordless, magnetically-attached design allows for easy, movable installation. Outdoor deployment avoids poles and cables—cameras can simply be tied to a tree. The low-power design also enables long battery life; one smart parking recorder product can last up to half a month on a single charge.
However, as a rapidly growing new brand, Yang revealed that Shenmou has attracted competitive attention and faced tactics like rival brands pressuring distributors for exclusivity. Shenmou's counter-strategy is twofold: first, pledging never to demand exclusivity from distributors as it grows, and second, currently supporting distributors in managing multiple brands simultaneously.
"We also sell our self-developed low-power chips to competitors," he said. "We hope for fair competition on a level playing field, driving the entire industry to develop the wireless low-power camera sector and promote industrial upgrading."
A Vision for Affordable, Ubiquitous Computing
Shenmou's product line now spans smart care, portable devices, vehicle monitoring, and smart terminals.
Yang draws a parallel to Steve Jobs' approach with the iPhone: first create the product, then the demand. With power and cost hurdles addressed, Shenmou adheres to first principles, contemplating user needs to introduce disruptive products.
He forecasts that as cameras transition to wireless, global annual shipments could hit 10 billion by 2035 and potentially 100 billion by 2045. Shenmou's market share is poised for explosive growth within this trend. With 2025 sales at 600 million yuan, he projects 2026 sales to reach 4 billion yuan, with targets of 20 billion and 100 billion yuan for 2027 and 2028, respectively.
In reality, Yang's ambitions extend beyond security cameras. With the explosive growth of AI glasses, smart robots, and massive IoT devices, the application scenarios for low-power chips hold vast potential. It is reported that the "Yanjixin" chip from parent company Hangzhou Yanjiwei Electronics is already used in Li AI glasses. Yang disclosed that "Yanjixin" could eventually find its way into smartphones, PCs, and even servers.
He emphasizes that AI is a crucial tool for humanity but must be affordable for widespread adoption. Currently, an average user's monthly electricity bill is around 100 yuan. For AI to be accessible to all, this cost needs to drop to about 30 yuan per month, or roughly one yuan per day—a problem he and Yanjiwei are striving to solve.
Besides Shenmou, Yanjiwei has also established a chip division—Shendeng, focused on edge-side AI inference, and a cloud computing division—Shensuan, handling chips and servers. In the chip business, he has set his sights on competing with NVIDIA.
In Yang's view, Chinese companies surpassing NVIDIA is not impossible. Currently, training compute is primarily used by large model companies, serving engineers. Even if there are 1,000 large model teams globally, each with 10,000 cards, that totals only 10 million cards. In contrast, demand for inference compute is far greater, as it serves the world's 6 billion people, and potentially 60 billion or even 600 billion AI agents in the future. Every recognition event triggers an inference request. As large models integrate into daily life and work, the cumulative consumption from inference will far exceed that of training.
He judges that while Chinese firms may find it difficult to surpass NVIDIA in training compute, they have a strong opportunity to lead in inference compute. This is because inference compute does not rely on the CUDA interface, allowing for extensive custom chip development—precisely where his fully-custom chip design methodology excels.
As he shared at a recent industry conference, computing power represents the fourth industrial revolution. With electricity demand rising in tandem with compute expansion, the "fully-custom" technical path is poised to ignite an efficiency revolution in "low-power" computing, making compute power as cheap and universally accessible as water and electricity. This, he concludes, is the ultimate vision for every tech company: to make technology both universally beneficial and benevolent.
Comments